Point-in-Time Data Explained for Stock Research
The two-clock model that prevents later filings, revisions, schedules, and identities from leaking into historical decisions.
Point-in-time data answers what value or event was eligible at a specific historical decision time. It keeps effective time—when something economically applies—separate from known-at time—when the information became observable. Revisions are appended with new known-at times, allowing an as-of query to exclude future corrections and reconstruct the original information set.
Every record needs the right clock
A fiscal quarter may end months before its 10-Q is filed. An insider trade occurs before its Form 4 becomes public. An earnings date can be estimated, revised, confirmed, and finally reported. One date cannot represent all these states.
Store source event time, public observation time, ingestion time, and version where relevant. Define which clock the strategy uses before joining datasets.
Revision history prevents backward travel
Restated fundamentals, amended filings, corrected news, and schedule changes should produce new versions. A latest-value table can still be offered for current analysis, but it must not be the source for historical as-of queries.
Resolve the eligible version only after applying the cutoff. Saving only the latest row makes faithful reconstruction impossible.
- Separate effective and known-at time.
- Append revisions.
- Use point-in-time security universes.
- Record the cutoff with every result.
Test the pipeline
Choose historical cutoffs around known amendments or schedule changes and verify that future versions disappear. Test after-hours events against the next exchange session.
DataCedar carries these clocks across filings, fundamentals, earnings, transcripts, news, and coverage so evidence can be joined without a different hidden timing rule for every stream.
How DataCedar preserves the evidence
DataCedar separates acquisition from serving. Permitted source responses are retained with retrieval time and identifiers, normalized into DataCedar-owned tables, checked against expected coverage, and exposed through a stable versioned API. A collector can be replaced without changing the customer contract or making an upstream provider a runtime dependency.
Every research stream carries effective and known-at time where the distinction matters. Rights-restricted, unavailable, partial, stale, and genuinely empty states remain visible, so a backtest can fail closed and a buyer can see the product boundary before committing engineering time.
Key takeaways
- 01Effective time is not public availability time.
- 02Revisions must be append-only versions.
- 03Apply the cutoff before selecting values.
- 04Test with real changes and amendments.
Query public company evidence through one stable API. Free Explorer tier, no card.
Start freeView pricing